[AA rework] implemetation of hashOf

[IgorStepanov]

This pull request is first stage of @dawgfoto associative array reworking strategy. This PR introduces hashOf function, which can compute hash value for different time (including in CTFE if possible). The next stage is implementing of new templated AssociativeArray struct without using TypeInfo.

Note: hash value, computed with hashOf not equal TypeInfo.getHash(...) hash value.

[MartinNowak]

I generated 4 random floats, can you add them to the unittest casted to float, double and real.

-0x9.0f7ee55df77618fp-13829
0x7.36e6e2640120d28p+8797
-0x1.05df6ce4702ccf8p+15835
0x9.54bb0d88806f714p-7088

[IgorStepanov]

done

[MartinNowak]

Thanks, look amazing.
Of course the floating point stuff is ridiculously complex.

[MartinNowak]

The hashOf function should be made accessible through object or some other core module.

@safe nothrow // pure can only be inferred because Object.toHash isn't pure
size_t hashOf(T)(auto ref T val)
{
    import core.internal.hash : hashOf;
    return hashOf(val);
}

I'd also like to mark all core.internal modules with package: protection but that doesn't work because object isn't in the core package.

[IgorStepanov]

The hashOf function should be made accessible through object or some other core module.

I think, there isn't reason to publish hashOf now. (BTW it conflits with rt.util.hash.hashOf).
It should be publised after creating new AA.

[IgorStepanov]

Thanks. The next stage is implementing new templated aa (as core.internal.aa). I hope, I'll do it during this week.

[MartinNowak]

This can be fairly simple. In rt.util.container.HashTab I've used the lower N bits for indexing, a singly linked list for each bucket and a simple grow policy to regulate the load factor.